Maoning Wang 2016-06-13
1. I cannot find (economic or mathematical) theoretical description of LiBin's concrete mean reversion assumption. It seems hard to do hypothesis testing.
2. The paper 'SEQUENTIAL MONTECARLO SAMPLING FOR CORRELATED LATENTLONG-MEMORY TIME-SERIES' (ICASSP 2016) studies how to generate random data following models like dS/S=udt+cdx, where dx follows normal distribution.
1.understand how to change the differential equation dS/S=udt+cdx into the example mentioned in the ICASSP 2016 paper.
2.understand how LiBin changes a learning algorithm into an OLP algorithm.